Vision in the production process

18 October 2011

Dr John Haddon, of the UK Industrial Vision Association, discusses the benefits that industrial vision system can offer at all stages of the production process

Industrial vision systems can introduce automation into the production process at a number of different levels – from speeding up the inspection process to levels not possible using human inspectors through to being an integral part of a statistical process control (SPC) system that can identify when a manufacturing process is moving out of specification.

Vision can offer fast, accurate and reproducible inspection capabilities at a competitive cost on finished products and can be linked to reject mechanisms to automatically ensure that no defect product reaches the end-user.

Using vision earlier in the production process can bring added value by improving quality throughout the process, reducing waste and improving process efficiency through energy saving. Traditionally, measurement data for SPC was acquired on a sampling basis, allowing trends in the process to be identified, which may then allow remedial action to be taken before the process starts to produce ‘out of spec’ product. However, the speed and accuracy of today’s industrial vision systems means that in many applications, 100% inspection can be carried out, and every product or component can be measured. By feeding this data into the SPC system trends can be identified earlier and random and sudden defects can be identified.

Vision and robotics
Industrial vision systems can provide powerful automation capabilities for industrial processes, but even more versatility is possible when they are combined with robotics. Image processing, for example, provides mechanical robots with a new mode of perception by giving them added ‘vision’, which is used to locate parts or guide the robot. In an alternative application, the robot is used to present parts for vision inspection. Both can have advantages over manual systems by automating transfers from one process to another or to feed parts into a system. Robot vision can simplify automation by reducing or eliminating the need for bespoke mechanical handling as well as being able to cope with many different parts within a single system. The emergence of high-speed 3D machine vision technologies enables automated high-speed pick and place robotics solutions for new and challenging applications, for example gripping of complex shapes and profiles, picking variable heights and picking products individually from a random arrangement from a variety of in-feeds.

Integrating vision
Vision systems come in a variety of configurations ranging from PC-based systems through to self-contained smart cameras with built-in image processing capabilities. The arrival of the GigE Vision and GeniCam standards in 2006 heralded a new dawn in machine vision technology by allowing hardware and software from different vendors to interoperate seamlessly using Ethernet technology to provide fast communication over distances up to 100m, with sufficient bandwidth on ‘normal’ network connections to meet the demands of most image processing applications. This allows the automation of a production line inspection process with versatile processing and control possibilities to be achieved using a ‘network centric’ approach.

With Ethernet products available for I/O, triggering and lighting control, all the elements required for component sensing, camera triggering and reject gate control needed to run a production-line vision system can be accessed using Gigabit Ethernet. Figure 2 shows a possible configuration for a conveyorised product line using a GigE Vision-based inspection system to generate a ‘pass/fail’ decision based on the measurements made, linked into a reject mechanism to create a fully automated inspection system.

Where physical space for mounting the camera is limited, a new breed of Power-over-Ethernet cameras is emerging where both power and data are sent over a single cable between the camera and the controlling PC, eliminating the separate power cable and its power supply.

For applications involving the integration of vision and robot systems, the situation can be more complex. In the past many users have been reluctant to integrate image processing into their robots because providing an interface to the camera and configuring the whole assembly was such a complex task, demanding expertise in both robotics and image processing. To overcome this, a number of companies have developed specific imaging interfaces for robot applications.

There are a number of important factors to consider when integrating a vision system into a manufacturing process – whether using an ‘off-the-shelf’ system or designing and creating the system from scratch. There needs to be a good match between the vision hardware and the goals of the final development. There is no point spending a fortune on a high-resolution colour camera with a fancy interface if a lower resolution mono camera will give sufficient data. Conversely, going to a low resolution camera with a poor software development kit and limited functionality may cause a development to be more difficult and time consuming than necessary and even to fail.

The development needs to be joined up – the development of the vision component needs to be done in a coherent way with other components, for example. It is also important to consider the potential trade-offs and the cost/benefits at an early stage of development. A small compromise on the algorithm might make a major difference in speed, accuracy and/or performance.

A practical example
A practical example illustrates both the range of challenges and the capabilities of vision integrated with a pick and place system. A pick-and-place quality control transfer system with integrated vision, quality control and reject facilities has been designed, manufactured and installed by a vision systems integrator for a medical device manufacturer to simplify its production process. The system was required to inspect foam pads pressed out from a roll of material and transfer them to a double conveyor belt system for further processing or rejection.

There are up to 10 different pad shapes manufactured and the pads are cut in patterns to minimise the uncut waste material. They are presented for inspection in arrays but with alternate rows in a different orientation. Equipped with two 4K resolution linescan cameras, the vision system is capable of detecting 0.5mm defects such as inclusions, contaminants and general manufacturing faults over an area of 0.5 x 0.7m as well as gathering geometric data for the pick calculations. The requirement is to select the pressed pads and transfer them to one of two conveyor belts at rates of up to 168 products per minute. One belt moves correctly orientated product to the next step of the process. The other can either rotate product by 180˚ and then send it forward or alternatively send it to a reject bin, depending on the instruction received from the vision system.

The system uses image data to send X, Y and Theta offset coordinates to the 4 servo motors controlling the pick head via an Ethernet connection to allow for random variations in the product position prior to picking. This ensures that the pick head is in its optimum pick position for each array. It was not practical to remove individual pads containing defects, so the system has to calculate an array size for rejection that contains the minimum amount of ‘good’ product in order to minimize waste. The system was capable of changing between any of the pad shape recipes in less than two minutes.

About the author
Dr. John Haddon is technical consultant to UK Industrial Vision Association (UKIVA) and director at Panther Vision Ltd

Contact Details and Archive...

Related Articles...

Print this page | E-mail this page